Robert Vaser

I am hiring a postdoc at the Genome Institute of Singapore. Join @NiranjanTW and me on a project aiming to rapidly identify microbes in a sample using sequencing data and deep learning methods inspired by song recognition. DM or email me

Scientists from GIS, namely Prof Mile Sikic and Postdoctoral Fellow Dr. Robert Vaser have published their work on a novel de novo genome assembler called 'Raven in Nature Computational Science'. Check it out! Read: [bit.ly/3faKGOW]

De novo assembler Raven is published at NatCompSci rdcu.be/ckZT1. Easy to use, fast and memory efficient. Try it with metagenomic data too. w/@robertvaser !!!

@robertvaser and @msikic present Raven, a graph drawing based method for improvement of genome assembly from erroneous long-reads (rdcu.be/ckZT1). nature.com/articles/s4358…

@UYunusbaev Unfortunately no, Raven produces a collapsed assembly.

@robertvaser Does Raven have keep-haplotype option?

We decreased memory requirements to ~40% of the input FASTA size plus a constant not exceeding 16GB in the latest Raven version (github.com/lbcb-sci/raven…), without any significant impact on execution time. 1/2

@cyxiang1027 @msikic RC:i:<int> is the number of reads concatenated into the contig. XO:i:<bool> denotes if the contig is circular (1 - is, 0 - is not). Circular contigs can also be seen in GFA.

@msikic @robertvaser SAM-like tag in sequence headers from Raven，what does "RC:i:8 XO:i:0" mean？How can i tell if the sequence is circular?

Raven version 1.5.0 - up to 2x faster (increases with data accuracy), improves NGA50 in most cases, properly handles small plasmids. w/ @robertvaser github.com/lbcb-sci/raven

glad that SWORD - fast and sensitive software for protein sequence alignment finally is gaining popularity. github.com/rvaser/sword academic.oup.com/bioinformatics… w/ @robertvaser and Dario Pavlovic

@lh3lh3 @IvanSovic @XLR I think I know where the bug is. As Ivan already mentioned, Racon retains the longest overlap per read, but the last longest overlap is retained, instead of the first. Thanks for testing/answers!

@robertvaser @IvanSovic @XLR best alignment first

A technical question (could save me a few hours on experiment): does Racon use all alignments in SAM for consensus or does it ignore secondary alignments?

@lh3lh3 @IvanSovic @XLR In which order does minimap2 output alignments of the same read?

@robertvaser @IvanSovic @XLR This is T2T CHM13. 3.05Gb in size. Alignments of the same query are always clustered together. Note that many reads are aligned in full length to multiple locations.

@lh3lh3 @IvanSovic @XLR Minimap2 transforms 4Gbp into minimizer index? Alignments per read are outputted in one block, i.e. the SAM is already sorted by read name/id?

@robertvaser @IvanSovic @XLR No options. Overlaps from immediate minimap2 SAM output. I believe racon is still using all the secondary alignments.

@lh3lh3 @IvanSovic @XLR Which parameters did you use for Racon?

@IvanSovic @XLR Did the experiment anyway. For T2T HiFi, --secondary=no reduces substitution errors from 379.3k to 3.4k, a 100-fold reduction.

🧬2nd edition of the course "GENOME ASSEMBLY USING @nanopore SEQUENCING" will be delivered online in February (22-26). 👤👤Still a few places available! 🔗physalia-courses.org/courses-worksh… #GenomeAssembly #Nanopore #longreads

Similar requirements hold for Racon as well, although this improvement is on branch library (github.com/lbcb-sci/racon…) which currently does not support overlaps/alignments from the outside. Any feedback is welcome! 2/2

Interested in #GenomeAssembly using @nanopore data? Still a few places on the second edition of our course with @robertvaser and @jmaricb (@msikic lab) in February (22-26). Please check it out: physalia-courses.org/courses-worksh…

Dedicate enough time to database preparation biorxiv.org/content/10.110… w/ @jmaricb @KrizanovicK @Sylvain14518009 @NiranjanTW

Registrations are now open for the 2nd edition of the course "GENOME ASSEMBLY USING OXFORD NANOPORE SEQUENCING" in February 2021 with @robertvaser and @jmaricb physalia-courses.org/courses-worksh… #GenomeAssembly #LongReads @nanopore @NanoporeConf

@andreaswallberg @msikic That depends on the sequencing depth. The memory consumptions should be qual to ~1.3 times the FASTQ file.

@msikic @robertvaser It looks great! I am trying it on a complex 18Gbp genome using Nanopore data, that was previously assembled with wtdbg2. Will probably not be able to fit it into the 1 TB RAM machine I have access to atm but hope to be able to run it on a 4.7 TB RAM later. Tips are welcome!

Raven - rapid and easy to use long read de novo assembler which produces the least number of contigs biorxiv.org/content/10.110… w/ @robertvaser. Any comments are more than welcome!